Hourglass-ShapeNetwork Based Semantic Segmentation for High Resolution Aerial Imagery
Open Access
- 25 May 2017
- journal article
- research article
- Published by MDPI AG in Remote Sensing
- Vol. 9 (6) , 522
Abstract
A new convolution neural network (CNN) architecture for semantic segmentation of high resolution aerial imagery is proposed in this paper. The proposed architecture follows an hourglass-shaped network (HSN) design being structured into encoding and decoding stages. By taking advantage of recent advances in CNN designs, we use the composed inception module to replace common convolutional layers, providing the network with multi-scale receptive areas with rich context. Additionally, in order to reduce spatial ambiguities in the up-sampling stage, skip connections with residual units are also employed to feed forward encoding-stage information directly to the decoder. Moreover, overlap inference is employed to alleviate boundary effects occurring when high resolution images are inferred from small-sized patches. Finally, we also propose a post-processing method based on weighted belief propagation to visually enhance the classification results. Extensive experiments based on the Vaihingen and Potsdam datasets demonstrate that the proposed architectures outperform three reference state-of-the-art network designs both numerically and visually.Keywords
Funding Information
- Fonds Wetenschappelijk Onderzoek (G084117)
- Brussels Institute for Research and Innovation (Innoviris) (3DLicornea)
This publication has 25 references indexed in Scilit:
- Salient Band Selection for Hyperspectral Image Classification via Manifold RankingIEEE Transactions on Neural Networks and Learning Systems, 2016
- Constructing the L2-Graph for Robust Subspace Learning and Subspace ClusteringIEEE Transactions on Cybernetics, 2016
- Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-scale Convolutional Architecture2015
- Hyperspectral Anomaly Detection by Graph Pixel SelectionIEEE Transactions on Cybernetics, 2015
- Features, Color Spaces, and Boosting: New Insights on Semantic Classification of Remote Sensing ImagesIEEE Transactions on Geoscience and Remote Sensing, 2014
- CNN Features Off-the-Shelf: An Astounding Baseline for Recognition2014
- Pooling in image representation: The visual codeword point of viewComputer Vision and Image Understanding, 2013
- A survey of multilinear subspace learning for tensor dataPattern Recognition, 2011
- Automatic Subspace Clustering of High Dimensional DataData Mining and Knowledge Discovery, 2005
- Factor graphs and the sum-product algorithmIEEE Transactions on Information Theory, 2001